Stock Analysis Based on the ARIMA Model

Abstract

This study applies the ARIMA model to analyze and forecast the daily opening price data of Industrial and Commercial Bank of China (stock code: 601398.SH), spanning from January 2019 to December 2024, with a total of 1,456 observations. The Augmented Dickey-Fuller (ADF) test confirms that the original series is non-stationary but becomes stationary after first-order differencing, establishing it as an I(1) process. The autocorrelation function (ACF) and partial autocorrelation function (PACF) plots of the differenced series exhibit a clear cutoff at lag 1, guiding model identification. Among 16 candidate ARIMA models evaluated using AIC, BIC, and HQIC criteria, the ARIMA (1, 1, 2) model is selected as optimal, achieving the lowest values across all three information criteria. Diagnostic checks, including the Ljung-Box Q-test at various lag orders (5, 10, 15, 18), indicate that residuals are white noise, confirming adequate model specification. The model demonstrates strong in-sample fit with an R-squared of 0.9145. Forecast evaluations over short-term (10 days), medium-term (50 days), and long-term (116 days) horizons yield MAE/RMSE values of 0.07/0.09, 0.05/0.06, and 0.08/0.11, respectively, indicating robust predictive performance across different horizons. The findings suggest that the ARIMA (1, 1, 2) model effectively captures the underlying dynamics of stock opening prices and provides reliable forecasts, offering valuable decision-making support for investors in volatile financial markets.